IEEE COMMUNICATIONS MAGAZINE 1 Index Modulation Techniques ... · IEEE COMMUNICATIONS MAGAZINE 1 Index Modulation Techniques for 5G Wireless Networks Ertugrul Basar, Senior Member,

IEEE COMMUNICATIONS MAGAZINE 1

Index Modulation Techniques for 5G Wireless NetworksErtugrul Basar, Senior Member, IEEE

Abstract

The increasing demand for higher data rates, better quality of service, fully mobile and connected wirelessnetworks lead the researchers to seek new solutions beyond 4G wireless systems. It is anticipated that 5G wirelessnetworks, which are expected to be introduced around 2020, will achieve ten times higher spectral and energyefficiency than current 4G wireless networks and will support data rates up to 10 Gbps for low mobility users. Theambitious goals set for 5G wireless networks require dramatic changes in the design of different layers for nextgeneration communications systems. Massive multiple-input multiple-output (MIMO) systems, filter bank multi-carrier (FBMC) modulation, relaying technologies, and millimeter-wave communications have been considered assome of the strong candidates for the physical layer design of 5G networks. In this article, we shed light on thepotential and implementation of index modulation (IM) techniques for MIMO and multi-carrier communicationssystems which are expected to be two of the key technologies for 5G systems. Specifically, we focus on twopromising applications of IM: spatial modulation (SM) and orthogonal frequency division multiplexing with IM(OFDM-IM), and we discuss the recent advances and future research directions in IM technologies towards spectraland energy-efficient 5G wireless networks.

I. INTRODUCTION

After more than 20 years of research and development, the achievable data rates of today’s cellularwireless communications systems are several thousands of times faster compared to earlier 2G wirelesssystems. However, unprecedented levels of spectral and energy efficiency are expected from 5G wirelessnetworks to achieve ubiquitous communications between anybody, anything, and anytime [1]. In order toreach the challenging objectives of 5G wireless networks, the researchers have envisioned novel physicallayer (PHY) concepts such as massive multiple-input multiple-output (MIMO) systems and non-orthogonalmulti-carrier communications schemes. However, the wireless community is still working day and nightto come up with new and more effective PHY solutions towards 5G networks. There has been a growinginterest on index modulation (IM) techniques over the past few years. IM, in which the indices of thebuilding blocks of the considered communications systems are used to convey additional information bits,is a novel digital modulation scheme with high spectral and energy efficiency. Spatial modulation (SM) andorthogonal frequency division multiplexing with IM (OFDM-IM) schemes, where the corresponding indexmodulated building blocks respectively are the transmit antennas of a MIMO system and the subcarriersof an OFDM system, appear as two interesting as well as promising applications of the IM concept.

After the pioneering works of Mesleh et al. [2] and Jeganathan et al. [3], SM techniques have attractedsignificant attention over the past few years. Although having strong and well-established competitors suchas vertical Bell Labs layered space-time (V-BLAST) and space-time coding (STC) systems, SM schemeshave been regarded as possible candidates for spectral and energy-efficient next generation MIMO systems.On the other hand, the researchers have started to explore the potential of IM concept for the subcarriersof OFDM systems in recent times and it has been shown that the OFDM-IM scheme [4] can provideattractive advantages over classical OFDM, which is an integral part of many current wireless standards.

The aim of this article is to present the basic principles of these two promising schemes, SM and OFDM-IM, which are still waiting to be explored by many experts, and review some of the recent interestingresults in IM techniques. Furthermore, we discuss the implementation scenarios of IM techniques for

E. Basar is with Istanbul Technical University, Faculty of Electrical and Electronics Engineering, 34469, Istanbul, Turkey.e-mail: [email protected] work is supported in part by the Scientific and Technological Research Council of Turkey (TUBITAK) under Grant no. 114E607.Copyright (c) 2016 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes

must be obtained from the IEEE by sending a request to [email protected].

arX

iv:1

604.

0831

5v1

[cs

.IT

] 2

8 A

pr 2

016


next generation wireless networks and outline possible future research directions. Particularly, we shiftour focus to generalized, enhanced, and quadrature IM schemes and the application of IM techniques formassive multi-user MIMO (MU-MIMO) and cooperative communications systems.

II. INDEX MODULATION FOR TRANSMIT ANTENNAS: SPATIAL MODULATION

SM is a novel way of transmitting information by means of the indices of the transmit antennas ofa MIMO system in addition to the conventional M -ary signal constellations. In contrast to conventionalMIMO schemes which rely either on spatial multiplexing to boost the data rate or spatial diversity toimprove the error performance, the multiple transmit antennas of a MIMO system are used for a differentpurpose in an SM scheme. More specifically, there are two information carrying units in SM: indices oftransmit antennas and M -ary constellation symbols. For each signaling interval, a total of

log2(nT ) + log2(M) (1)

bits enter the transmitter of an SM system as seen from Fig. 1, where nT and nR denote the numberof transmit and receive antennas, respectively, and M is the size of the considered signal constellationsuch as M -ary phase shift keying (M -PSK) or M -ary quadrature amplitude modulation (M -QAM). Thelog2(M) bits of the incoming bit sequence are used to modulate the phase and/or amplitude of a carriersignal traditionally, while the remaining log2(nT ) bits of the incoming bit sequence are reserved forthe selection of the index (I) of the active transmit antenna which performs the transmission of thecorresponding modulated signal (s).

The receiver of the SM scheme has two major tasks to accomplish: detection of the active transmitantenna for the demodulation of the index selecting bits and detection of the data symbol transmitted overthe activated transmit antenna for the demodulation of the bits mapped to M -ary signal constellation.Unfortunately, the optimum maximum likelihood (ML) detector of SM has to make a joint search overall transmit antennas and constellation symbols to perform these two tasks. In other words, the MLdetector of the SM scheme independently implements a classical single-input multiple-output (SIMO)ML detector for all transmit antennas to find the activated transmit antenna by comparing the minimumdecision metrics (m1,m2, . . . ,mnT

). On the other hand, the primitive suboptimal detector of SM dealswith the aforementioned two tasks one by one, that is, first, it determines the activated transmit antenna,second, it finds the data symbol transmitted over this antenna. Therefore, the size of the search spacebecomes nT ×M and nT +M for the ML and suboptimal detector, respectively.

SM systems provide important advantages over classical MIMO systems, which are extensively coveredin the literature [5], [6]. The main advantages of SM over classical MIMO systems can be summarizedas follows:

• Simple transceiver design: Since only a single transmit antenna is activated, a single radio frequency(RF) chain can handle the transmission for the SM scheme. Meanwhile, inter-antenna synchronization(IAS) and inter-channel interference (ICI) are completely eliminated, and the decoding complexityof the receiver, in terms of total number of real multiplications performed, grows linearly with theconstellation size and number of transmit antennas.

• Operation with flexible MIMO systems: SM does not restrict the number of receive antennas as theV-BLAST scheme, which requires nR > nT to operate with minimum mean square error (MMSE)detector.

• High spectral efficiency: Due to the use of antenna indices as an additional source of information,the spectral efficiency of SM is higher than that of single-input single-output (SISO) and orthogonalSTC systems.

• High energy efficiency: The power consumed by the SM transmitter is independent from number oftransmit antennas while information can be still transfered via these antennas. Therefore, SM appearsas a green and energy-efficient MIMO technology.


𝑀𝑀 - ary Modulation

Antenna Indices Selection

SM

Mapper

𝑠𝑠

𝐼𝐼 1

2

𝑛𝑛𝑇𝑇

log2(𝑀𝑀) bits

log2(𝑛𝑛𝑇𝑇) bits

SM ML Detector

1

𝑛𝑛𝑅𝑅

C H A N N E L

SIMO 1

SIMO 2

SIMO 𝑛𝑛𝑇𝑇

2

⋮ ⋮ ⋮

Channel Estimation ⋮

��𝑠,𝑚𝑚1

��𝑠,𝑚𝑚2

��𝑠,𝑚𝑚𝑛𝑛𝑇𝑇

Min Selector

��𝑠, 𝐼𝐼

SM Demapper

��𝑠, 𝐼𝐼

log2(𝑀𝑀) bits

log2(𝑛𝑛𝑇𝑇) bits

⋮

Fig. 1. Block diagram of the SM transceiver for an nT × nR MIMO system. s (or s) and I (or I) ∈ {1, 2, . . . , nT } denote the selected (orestimated) M -ary constellation symbol and transmit antenna index, respectively and mn, n = 1, 2, . . . , nT is the minimum metric providedby the nth SIMO ML detector.

As an example, assuming an nT × nR MIMO system operating at a fixed spectral efficiency, SMachieves 200(nT −1)/(2nT +1)% reduction in ML detection complexity (in terms of total number of realmultiplications) compared to V-BLAST due to the activation of a single transmit antenna. Furthermore,the sparse structure of the transmission vectors allows the implementation of several near/sub-optimallow-complexity detection methods for SM systems such as matched filter based detection and compressedsensing based detection. In terms of the energy efficiency in Mbits/J, improvements up to 46% comparedto V-BLAST are reported for different type of base stations (BSs) equipped with multiple antennas.

While the SM scheme has the aforementioned appealing advantages, it also has some disadvantages.The spectral efficiency of SM increases logarithmically with nT , while the spectral efficiency of V-BLASTincreases linearly with nT . Therefore, higher number of transmit antennas are required for SM to reachthe same spectral efficiency as that of V-BLAST. The channel coefficients of different transmit antennasmust be sufficiently different for an SM scheme to operate effectively. In other words, SM requires richscattering environments to ensure better error performance. Since SM transfers the information using onlythe spatial domain, plain SM cannot provide transmit diversity as STC systems which rely on both spatialand time domains for data transmission.

Considering the advantages and disadvantages of SM systems, we may conclude that SM provides aninteresting trade-off among complexity, spectral efficiency, and error performance. Consequently, SMhas been regarded as a possible candidate for spectral and energy-efficient next generation wirelesscommunications systems [1].

III. RECENT ADVANCES IN SMThe first studies on SM concept date back to the beginning of 2000s in which the researchers used

different terminologies. However, after the inspiring works of Mesleh et al. [2] and Jeganathan et al.[3], numerous papers on SM have been published in which the experts focus on generalized, spectraland energy-efficient SM systems, low-complexity detector types, block/trellis coded SM systems withtransmit/time diversity, link adaptation methods such as adaptive modulation, transmit antenna selectionand precoding, performance analysis for different fading channel types and channel estimation errors,


information theoretical analyses, differential SM schemes with non-coherent detection, cooperative SMsystems, and so on. For a comprehensive overview of these studies, interested readers are referred tosurvey papers [5] and [6].

In this section, we review some of the recent as well as promising advances in SM technologies such asgeneralized, enhanced, and quadrature SM systems, massive MU-MIMO systems with SM, and cooperativeSM schemes, which have the potential to provide efficient solutions towards 5G wireless networks.

A. Generalized, Enhanced, and Quadrature SM SchemesAs mentioned earlier, the major disadvantage of SM is its lower spectral efficiency compared to

classical V-BLAST scheme for the same number of transmit antennas. Although a considerable numberof information bits can still be transmitted by the indices of active transmit antennas, for higher ordermodulations and MIMO systems, SM suffers a significant loss in spectral efficiency with respect to V-BLAST due to its inactive transmit antennas.

One of the first attempts to not only increase the spectral efficiency of SM but also ease the constrainton number of transmit antennas, which has to be an integer power of two for classical SM, has been madein [7] by the generalized SM (GSM) scheme, where the number of active transmit antennas is no longerfixed to unity. In the GSM scheme, multiple active transmit antennas are selected to transmit the same datasymbol. Denoting the number of active transmit antennas by nA where nA < nT , blog2

(nT

nA

)c information

bits can be transmitted for each signaling interval in addition to the log2(M) bits transmitted by the M -arydata symbols, where b·c is the floor operation. Since log2(nT ) ≤ blog2

(nT

nA

)c for nT = 2n (n = 1, 2, . . .),

the spatial domain can be used more effectively by the GSM scheme. As an example, for nT = 8, onlythree bits can be transmitted by the antenna indices in SM, while this can be doubled by GSM for nA = 4.In [8], the concept of GSM has been extended to multiple-active spatial modulation (MA-SM), wheredifferent data symbols are transmitted from the selected transmit antennas to further boost the spectralefficiency. Therefore, the spectral efficiency of the MA-SM scheme becomes blog2

(nT

nA

)c + nA log2(M)

bits per channel use (bpcu), which is considerably higher than that of SM. It should be noted that MA-SMprovides an intermediate solution between two extreme schemes: SM and V-BLAST, which are specialcases of MA-SM with nA = 1 and nA = nT , respectively.

Enhanced SM (ESM) is a recently proposed and promising variant of SM [9]. In the ESM scheme, thenumber of active transmit antennas can vary for each signaling interval and the information is conveyednot only by the indices of active transmit antennas but also by the selected signal constellations used intransmission. In other words, the ESM scheme considers multiple signal constellations and the informationis transmitted by the combination of active transmit antennas and signal constellations. As an example, fortwo transmit antennas and four bpcu transmission, the ESM scheme transmits two bits by the joint selectionof active transmit antennas and signal constellations, where one quadrature PSK (QPSK) and two binaryPSK (BPSK) signal constellations (one ordinary and one rotated) can be used [9]. For two-bit sequences{0, 0}, {0, 1}, {1, 0}, and {1, 1}, the ESM scheme uses the following transmission vectors, respectively:[S4 0

]T,[0 S4

]T,[S2 S2

]T, and[S2ejθ S2ejθ

]T, where Sm,m = 2, 4 denotes M -PSK constellation, (·)T

stand for transposition of a vector and θ = π/2 is a rotation angle used to obtain a third signal constellationin addition to classical BPSK and QPSK signal constellations. Other implementation examples of ESM canbe found in [9]. It is interesting to that the first two transmission vectors of the ESM scheme correspond tothe classical SM using QPSK with single activated transmit antenna, where the first and second transmitantenna is used for the transmission of a QPSK symbol, respectively. On the other hand, the third andfourth transmission vectors correspond to the simultaneous transmission of two symbols selected fromBPSK and modified BPSK constellations, respectively. The reason behind reducing the constellation sizefrom four to two can be explained by the fact that same number of information bits (two bits for thiscase) must be carried with M -ary constellations independent from number of active transmit antennas.Examples of the generalization of the ESM scheme for different number of transmit antennas and signalconstellations can be found in [9].


TABLE ITRANSMISSION VECTORS

(xT)

OF SM, ESM, AND QSM SCHEMES FOR 4 BPCU AND TWO TRANSMIT ANTENNAS (nT = 2), RED BITSINDICATE THE SINGLE BIT TRANSMITTED BY THE SPATIAL DOMAIN FOR SM, BLUE BITS INDICATE THE ADDITIONAL ONE BIT

TRANSMITTED BY THE SPATIAL DOMAIN FOR ESM AND QSM

Bits SM ESM QSM Bits SM ESM QSM

0000[1 0

] [1+j√

20] [

1+j√2

0]

1000[0 1

] [1√2

1√2

] [j√2

1√2

]0001

[1+j√

20] [

−1+j√2

0] [

−1+j√2

0]

1001[0 1+j√

2

] [1√2

−1√2

] [j√2

−1√2

]0010

[j 0] [

−1−j√2

0] [

−1−j√2

0]

1010[0 j] [

−1√2

1√2

] [−j√2

−1√2

]0011

[−1+j√

20] [

1−j√2

0] [

1−j√2

0]

1011[0 −1+j√

2

] [−1√2

−1√2

] [−j√2

1√2

]0100

[−1 0

] [0 1+j√

2

] [1√2

j√2

]1100

[0 −1

] [j√2

j√2

] [0 1+j√

2

]0101

[−1−j√

20] [

0 −1+j√2

] [−1√2

j√2

]1101

[0 −1−j√

2

] [j√2

−j√2

] [0 −1+j√

2

]0110

[−j 0

] [0 −1−j√

2

] [−1√2

−j√2

]1110

[0 −j

] [−j√2

j√2

] [0 −1−j√

2

]0111

[1−j√

20] [

0 1−j√2

] [1√2

−j√2

]1111

[0 1−j√

2

] [−j√2

−j√2

] [0 1−j√

2

]

Quadrature SM (QSM) [10] is yet another clever modification of classical SM to improve the spec-tral efficiency while maintaining its advantages such as operation with single RF chain and ICI freetransmission. In the QSM scheme, the real and imaginary parts of the complex M -ary data symbolsare separately transmitted using the SM principle. For a MIMO system with nT transmit antennas, thespectral efficiency of QSM becomes 2 log2(nT )+ log2(M) bpcu by simultaneously applying SM principlefor in-phase and quadrature components of the complex data symbols. As an example, for nT = 2 andM = 4, in addition to the two bits mapped to the QPSK constellation, extra two bits can be transmittedin the spatial domain by using one of the following four transmission vectors:

[sR + jsI 0

]T,[sR jsI

]T,[jsI sR

]T, and[0 sR + jsI

]T for input bit sequences {0, 0}, {0, 1}, {1, 0}, and {1, 1}, respectively, wheresR and sI denote the real and imaginary parts of s = sR + jsI ∈ S4 , respectively. It is interesting tonote that the first and second element of these two-bit sequences indicates the position of the real andimaginary part of s, respectively. Even if the number of active transmit antennas can be one or two forthe QSM scheme, a single RF chain is sufficient at the transmitter since only two carriers (cosine andsine) are used during transmission.

In Table I, transmission vectors of SM, ESM, and QSM schemes are given for 4 bpcu transmission andtwo transmit antennas, where we considered natural bit mapping for ease of presentation. We observe fromTable I that both ESM and QSM schemes convey more bits by the spatial domain compared to conventionalSM, which leads to not only improved spectral efficiency but also to higher energy efficiency.

In Fig. 2, we compare the minimum squared Euclidean distance between the transmission vectors(dmin), which is an important design parameter for quasi-static Rayleigh fading channels to optimize theerror performance, of SIMO, SM, ESM, and QSM schemes. In all considered configurations, the averagetotal transmitted energy is normalized to unity to make fair comparisons. It is interesting to note thatESM and QSM schemes achieve the same dmin value for 4 and 6 bpcu transmissions. However, as seenfrom Fig. 2, QSM suffers a worse minimum Euclidean distance, as a result a worse error performance,compared to ESM scheme for higher spectral efficiency values, while the ESM scheme requires a morecomplicated transmitter with two RF chains. Finally, the results of Fig. 2 also prove that the relative dminadvantage of IM schemes over classical SIMO scheme increases with increasing spectral efficiency, thatis, IM techniques become more preferable for higher spectral efficiency values.


(a) 4 bpcu, nT = 2

SIMO SM ESM QSM

dm

in

0

0.2

0.4

0.6

0.8

1

1.2

(b) 6 bpcu, nT = 4

SIMO SM ESM QSM

dm

in

0

0.2

0.4

0.6

0.8

1

1.2

(c) 8 bpcu, nT = 4

SIMO SM ESM QSM

dm

in

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

(d) 10 bpcu, nT = 4

SIMO SM ESM QSM

dm

in

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

Fig. 2. Minimum squared Euclidean distance (dmin) comparison of SIMO, SM, ESM and QSM schemes for different configurations (a)4 bpcu, nT = 2. SIMO:16-QAM, SM:8-PSK, ESM:QPSK/BPSK, QSM:QPSK. (b) 6 bpcu, nT = 4. SIMO:64-QAM, SM:16-QAM,ESM:QPSK/BPSK, QSM:QPSK. (c) 8 bpcu, nT = 4. SIMO:256-QAM, SM:64-QAM, ESM:16-QAM/QPSK, QSM:16-QAM. (d) 10 bpcu,nT = 4. SIMO:1024-QAM, SM:256-QAM, ESM:64-QAM/8-QAM, QSM:64-QAM.

B. Massive Multi-user MIMO Systems with SMMassive MIMO concept, in which the BSs have tens to hundreds of antennas, is considered as one

of the potential key technologies for 5G wireless networks due to its appealing advantages such as veryhigh spectral and energy efficiency. While the initial studies on MIMO systems generally focus on point-to-point links where two users communicate with each other, practical MU-MIMO systems are gainingmore and more attention to exploit the multiple antennas of a MIMO system to support multiple userssimultaneously.

The extension of MIMO systems into massive scale provides unique opportunities for SM systems sinceit becomes possible to transmit higher number of information bits by the spatial domain with massiveMIMO systems, even if the number of available RF chains is very limited. Although the spectral efficiencyof SM systems cannot compete with that of traditional methods such as V-BLAST for massive MIMOsystems, the use of IM concept for the transmit antennas of a massive MIMO system can provide an easyas well as cheap implementation solution thanks to the inherently available advantages of SM systems.Furthermore, SM is well-suited to unbalanced massive MIMO configurations in which the number ofreceive antennas are fewer than the number of transmit antennas [11].

In Fig. 3(a), a massive MU-MIMO system is considered where K users employ SM techniques foruplink transmission. Compared to user terminals with single antennas, additional information bits canbe transmitted using SM without increasing the system complexity. GSM, ESM, and QSM techniques


User 1

BS 𝑛𝑛𝑅𝑅~

10− 100

User 1 Data ⋮

1

𝑛𝑛𝑇𝑇1

User 2

User 2 Data

1

𝑛𝑛𝑇𝑇2

⋮

User 𝐾𝐾

User 𝐾𝐾 Data

1

𝑛𝑛𝑇𝑇𝐾𝐾

⋮

⋮

1

𝑛𝑛𝑅𝑅

BS 𝑛𝑛𝑇𝑇~

10− 100

1

𝑛𝑛𝑇𝑇

User 1 Data

User 𝐾𝐾 Data

⋮ User 2 Data ⋮

User 1

⋮

User 2

User 𝐾𝐾

⋮

1

𝑛𝑛𝑅𝑅1

⋮ 1

𝑛𝑛𝑅𝑅2

⋮ 1

𝑛𝑛𝑅𝑅𝐾𝐾

⋮

(a) (b)

Fig. 3. Massive MU-MIMO systems with SM (a) An uplink transmission scenario where User k has nkT transmit antennas available for SM

and the BS has nR ∼ 10− 100 receive antennas, (b) A downlink transmission scenario where User k has nkR receive antennas and the BS

has nT ∼ 10− 100 transmit antennas available for SM.

can be implemented at the users to further improve the spectral efficiency. At the BS, the optimal (ML)detector can be used at the expense of exponentially increasing decoding complexity (with respect to K)due to the inter-user interference. Low complexity near-optimal detection methods can be implementedas well by sacrificing the optimum error performance. Alternatively, SM techniques can also be used atthe BS for downlink transmission as shown in Fig. 3(b). To support high number of users, the massiveantennas of BS can be split into subgroups of fewer antennas where SM techniques can be employedfor each user [12]. For the specific case of two users, the data of User 1 can be mapped into antennaindices while the data of User 2 can be conveyed with M -ary signal constellations. GSM techniques canalso be implemented at the BS to transmit the data of different users with either antenna indices and/orconstellation symbols.

C. Cooperative SM SystemsCooperative communications, which allows the transmission of a user’s data not only by its own antenna,

but also by the active or passive nodes available in the network, has been one of the hot topics in thewireless communications field in the past decade. Initially, cooperative communication systems have beenproposed to create virtual MIMO systems for the mobile terminals due to the problems such as cost andhardware associated with the employment of multiple antennas in mobile terminals. However, due to therecent technological advances, multiple antennas can be employed at mobile terminals, and cooperativecommunications systems can efficiently provide additional diversity gains and high data rates by improvingcoverage. Consequently, relaying technologies have been incorporated into Long Term Evolution Advanced(LTE-A) standard for the purpose of increasing coverage, data rate, and cell-edge performance.

Considering the effective solutions provided by SM techniques and cooperative communications sys-tems, the combination of these two technologies naturally arises as a potential candidate for futurewireless networks. Due to the recent technological advances, cooperative SM systems can provide newimplementation scenarios, additional diversity gains, and higher data rates without increasing the cost andcomplexity of the mobile and relay terminals. In the past few years, researchers have shown that SMtechniques can be efficiently implemented for decode-and-forward (DF) and amplify-and-forward (AF)relaying based cooperative networks, distributed cooperation, and network coding systems. Readers arereferred to [5] and [13] and the references therein for further information on cooperative SM systems.


S ⋮

1

R

⋮

(a)

D

(d)

⋮ 1

𝑛𝑛𝐷𝐷

(b)

(c)

1st phase 2

nd phase

1 ... 𝑛𝑛𝑅𝑅

𝑛𝑛𝑆𝑆

R


S ⋮

1

𝑛𝑛𝑆𝑆

D ⋮

1

𝑛𝑛𝐷𝐷 D

⋮ 1

𝑛𝑛𝐷𝐷

R


S ⋮

1

𝑛𝑛𝑆𝑆

S ⋮

1

𝑛𝑛𝑆𝑆 D

⋮ 1

𝑛𝑛𝐷𝐷

R1


RN


Fig. 4. An overview of cooperative SM systems (a) Dual-hop SM (b) Cooperative SM (c) Network-coded SM (d) Multi-relay and distributedSM. nS , nR and nD denote the number of antennas for source (S), relay (R) and destination (D) nodes, respectively.

In Fig. 4, we consider four different cooperative SM system configurations where S, R, and D respec-tively stand for the source, relay, and destination node. In Fig. 4(a), a dual-hop network is considered,where SM techniques can be implemented at S and R with DF or AF based relaying techniques. Thescenario of Fig. 4(a) is generally observed in practical networks where S and D cannot communicatedirectly due to distance or obstacles. In this relaying scenario, SM techniques can improve the energyand spectral efficiency of S compared to single antenna case, while multiple RF chains are required at Rand D for signal reception. However, considering the uplink transmission from S to D, this would not bea major design problem. In Fig. 4(b), a direct link from S to D exists and R can improve the quality ofservice of the transmission between S and D by employing different relaying methods.

In Fig. 4(c), we take into account the two-way communications of S and D which is accomplished viaR. Without network coding, the overall transmission between S and D requires four transmission phases(from S to R, R to D, D to R, and R to S) which considerably reduce the overall spectral efficiency.However, the two-way communications between S and D can be accomplished at two phases with networkcoding where in the first transmission phase, S and D simultaneously transmit their signals to R usingSM techniques. In the second transmission phase, R combines the signals received from S and D, thenforwards this combined signal to S and D. The use of SM provides some opportunities for R such astransmitting one user’s data with antenna indices and the other one’s with constellation symbols.

Finally, in Fig. 4(d), we consider a distributed cooperation scenario with N relay nodes (R1, . . . ,RN ).In the first transmission phase, S can use SM techniques to transfer its data to relays. In the secondtransmission phase, one or more relays cooperate and the indices of the activated relays can be consideredas an additional way to convey information. This flexibility allows the relays to cooperate even if theyhave single antennas (nR = 1). Furthermore, opportunistic relay selection is also an option for thenetwork topology of Fig. 4(d), where the adaptively selected best relay takes part in transmission. For all


cooperation scenarios described above, S and/or R can use GSM/ESM/QSM techniques to further improvethe spectral efficiency as well as to obtain more flexibility in the system design.

IV. INDEX MODULATION FOR OFDM SUBCARRIERS: OFDM WITH INDEX MODULATION

IM concept can be considered for other communications systems apart from MIMO systems. For aninstance, IM techniques can be efficiently implemented for the subcarriers of an OFDM system. OFDM-IMis a novel multi-carrier transmission scheme which has been proposed by inspiring from the IM conceptof SM [4]. Similar to SM, in the OFDM-IM scheme, the incoming bit stream is split into index selectionand M -ary constellation bits. According to the index selection bits, only a subset of available subcarriersare selected as active, while the remaining inactive subcarriers are not used and set to zero. On the otherhand, the active subcarriers are modulated according to the M -ary constellation bits. In other words, theinformation is conveyed not only by the data symbols as in classical OFDM, but also by the indicesof the active subcarriers which are used for the transmission of the corresponding data symbols for theOFDM-IM scheme.

Considering an OFDM system with NF subcarriers, one can directly select the indices of activesubcarriers similar to IM technique used for the transmit antennas of an MA-SM system. However, themassive OFDM frames can provide more flexibility for the employment of IM techniques for OFDM-IMschemes compared to SM schemes. On the other hand, keeping in mind that NF can take very largevalues, such as 512, 1024 or 2048 as in LTE-A standard, there could be trillions of (actually more thana googol (10100) in mathematical terms) possible combinations for active subcarriers if index selection isapplied directly. As an example, assume that we want to select the indices of 256 active subcarriers outof NF = 512 available subcarriers, then, there could be 472.55× 10150 possible different combinations ofactive subcarriers, which turn the selection of active subcarriers into an almost impossible task. Therefore,for the implementation of OFDM-IM, the single and massive OFDM-IM block should be divided intoG smaller and manageable OFDM-IM subblocks each containing N subcarriers to perform IM, whereNF = G ×N . For each subblock, K out of N available subcarriers can be selected as active accordingto the p1 = blog2

(NK

)c index selection bits where typical N values could be 2, 4, 8, 16, and 32 with

1 ≤ K < N . Please note that classical OFDM becomes a special case of OFDM-IM with K = N , thatis, when all subcarriers are activated.

The block diagrams of OFDM-IM scheme’s transmitter and receiver structures are illustrated in Figs.5(a) and 5(b), respectively. As seen from Fig. 5(a), for each OFDM-IM frame, a total of

m = pG =

(⌊log2

(N

K

)⌋+K log2M

)G (2)

bits can be transmitted where p = p1+ p2 and p2 = K log2M . In Fig. 5(a), jg and sg denote the vector ofselected indices and M -ary data symbols with dimensions K×1, respectively. First, OFDM-IM subblockcreator forms the N×1 OFDM-IM subblocks xg, g = 1, . . . , G, then the OFDM-IM block creator obtainsthe NF × 1 main OFDM-IM frame x by concatenating these G OFDM-IM subblocks. After this point,G×N block interleaving can be performed to ensure that the subcarriers of a subblock are affected byuncorrelated wireless fading channels. Finally, classical OFDM procedures such as inverse fast Fouriertransform (IFFT), cyclic prefix (CP) insertion, and digital-to-analog (DAC) conversion are applied.

Two different index selection procedures are available for OFDM-IM: reference look-up tables forsmaller N values and combinatorial number theory method for higher N values, where examples of thesetwo methods are provided in Fig. 5(c). Similar to SM, the receiver of OFDM-IM has to determine theactive subcarriers and the corresponding data symbols in accordance with the index selection procedureused at the transmitter. After applying inverse operations, first, the received signals are separated sincethe detection of different subblocks can be carried out independently. The optimum but high-complexityML detector makes a joint search over possible subcarrier activation combinations and data symbols,while the low-complexity log-likelihood ratio (LLR) calculation based near-optimal detector determines


𝑚𝑚 bits

𝑝𝑝1 bits

𝑝𝑝2 bits

𝑝𝑝 bits

𝑝𝑝 bits

𝑀𝑀 - ary Mod.

Index Selection

⋮

𝑝𝑝1 bits

𝑝𝑝2 bits

𝑀𝑀 - ary Mod.

Index Selection

OFDM-IM

Subblock Creator

1

OFDM-IM

Subblock Creator 𝐺𝐺

OFDM-IM

Block Creator

𝐺𝐺 ×𝑁𝑁 Block

Interleaver 𝑁𝑁𝐹𝐹 - IFFT

Cyclic Prefix &

DAC ⋮

𝐣𝐣1

𝐣𝐣𝐺𝐺

𝐬𝐬1

𝐬𝐬𝐺𝐺

𝐱𝐱1

𝐱𝐱𝐺𝐺

𝐱𝐱 𝐱𝐱�

𝐪𝐪�

(a)

Cyclic Prefix &

ADC 𝑁𝑁𝐹𝐹 - FFT

𝐺𝐺 × 𝑁𝑁 Block

Deinter.

𝐲𝐲�

𝐲𝐲

𝐮𝐮�

Received Signals

Seperation

𝐲𝐲1

𝐲𝐲𝐺𝐺

ML / LLR Detector

𝑚𝑚 bits

(b)

𝑝𝑝1 bits

Index Selection

𝐣𝐣𝑔𝑔

𝑁𝑁 = 4,𝐾𝐾 = 2,�4

2� = 6⇒ 𝑝𝑝1 = 2 bits

{00}, {01}, {10}, {11} [1 3]T, [2 4]T, [1 4]T, [2 3]T Reference Look-up Table

𝑁𝑁 = 32,𝐾𝐾 = 16,�32

16� = 601, 080,390⇒ 𝑝𝑝1 = 29 bits

{0 1 0 0 0 0 1 1 0 1 0 0 0 1 1 0 0 0 1 1 1 0 1 1 0 0 0 1 0}

[29 28 27 26 25 19 17 16 15 14 13 12 11 8 4 2]T

Combinatorial Number Theory

(c)

bits indices

Fig. 5. OFDM-IM system at a glance (a) Transmitter structure (b) Receiver structure (c) Two different index selection procedures.

the indices of the active subcarriers first, then, it detects the corresponding data symbols. The LLR detectorcalculates a probabilistic measure on the active status of a given subcarrier by considering the fact that thecorresponding subcarrier can be either active (carrying an M -ary constellation symbol) or inactive. Thisdetector is classified as near-optimal since it does not know the set of all possible subcarrier activationcombinations.

It has been shown that OFDM-IM provides an interesting trade-off between error performance andspectral efficiency, and it offers some attractive advantages over classical OFDM. Unlike classical OFDM,the number of active subcarriers of an OFDM-IM scheme can be adjusted accordingly to reach the desired


spectral efficiency and/or error performance. Furthermore, due to the information bits carried by IM, whichhave lower error probability compared to ordinary M -ary constellation bits, OFDM-IM can provide betterbit error rate (BER) performance than classical OFDM for low-to-mid spectral efficiency values while itexhibits comparable decoding complexity using the near-optimal LLR detector. Furthermore, it has beenrecently proved that OFDM-IM also outperforms the classical OFDM in terms of ergodic achievable rate[14].

Consequently, we conclude that due to its flexible system design with adjustable number of activesubcarriers and its attractive advantages over OFDM, OFDM-IM can be a possible candidate not only forhigh-speed wireless communications systems but also for machine-to-machine (M2M) communicationssystems of 5G wireless networks which require low power consumption.

V. RECENT ADVANCES IN OFDM-IMSubcarrier IM concept for OFDM has attracted significant attention from the researchers in recent

times and it has been investigated in some up-to-date studies which deal with the error performanceand capacity analysis, generalization, enhancement, and optimization of OFDM-IM, and its adaptation todifferent wireless environments. Interested readers are referred to [15] and the references therein for anoverview of these studies. In this section, we focus on two recently proposed and promising forms ofOFDM-IM: generalized OFDM-IM and MIMO-OFDM-IM systems.

A. Generalized OFDM-IM SchemesTwo generalized OFDM-IM structures (OFDM-GIM-I and OFDM-GIM-II) have been recently proposed

by modifying the original OFDM-IM scheme [16]. In OFDM-GIM-I scheme, the number of active subcar-riers are no longer fixed and it is also determined according to the information bits. Considering the case ofN = 4, K = 2 with binary PSK (BPSK) modulation (M = 2), according to (2), blog2

(42

)c+2 log2(M) = 4

bits can be transmitted per OFDM-IM subblock, that is, a total of 4× 22 = 16 subblock realizations canbe obtained. On the other hand, considering all activation patterns (K ∈ {0, 1, 2, 3, 4}), which means thatthe number of active subcarriers can take values from zero (all subcarriers are inactive, K = 0) to four(all subcarriers are active, K = 4), as well as considering all possible values of M -ary data symbols, atotal of

∑NK=0

(NK

)MK = 81 possible subblock realizations can be obtained for which blog2(81)c = 6

bits can be transmitted per OFDM-GIM-I subblock. As a result, the OFDM-GIM-I scheme can providemore flexibility for the selection of active subcarriers and can transmit more bits per subblock comparedto OFDM-IM.

The OFDM-GIM-II scheme aims to further improve the spectral efficiency by applying IM inde-pendently for in-phase and quadrature components of the complex data symbols similar to the QSMscheme. In other words, a subcarrier can be active for one component, while it can be inactive simul-taneously for the other component. Considering the case of N = 16, K = 10 with quadrature PSK(QPSK) modulation (M = 4), according to (2), blog2

(1610

)c + 10 log2(M) = 32 bits can be transmitted

per OFDM-IM subblock. On the other hand, the OFDM-GIM-II scheme allows the transmission ofblog2

((1610

)(√M)K ×

(1610

)(√M)K

)c = 44 bits per subblock, which is 37.5% higher than that of OFDM-

IM. Please note that the in-phase and quadrature components of a complex M -QAM symbol are theelements of a

√M -ary pulse amplitude modulation (PAM) constellation, where a total of

(NK

)× (√M)K

realizations are possible per each component.

B. From SISO-OFDM-IM to MIMO-OFDM-IMThe first studies on OFDM-IM generally focused on point-to-point SISO systems, which can be

unsuitable for some applications due to their limited spectral efficiency. More recently, MIMO trans-mission and OFDM-IM principles are combined to further boost the spectral and energy efficiency of theOFDM-IM scheme [15]. Specifically, the transmitter of the MIMO-OFDM-IM scheme consists of parallel


SNR(dB)

0 5 10 15 20 25 30 35 40 45

BE

R

10-6

10-5

10-4

10-3

10-2

10-1

100

Classical MIMO-OFDM,2x2,1.94 bits/s/Hz



MIMO-OFDM-IM,2x2,1.94 bits/s/Hz



Fig. 6. Uncoded BER performance of MIMO-OFDM-IM and classical MIMO-OFDM schemes for three nT × nR MIMO configurations:2×2, 4×4, and 8×8. OFDM system parameters: M = 2 (BPSK), N = 4,K = 2, NF = 512, CP length = 16, frequency-selective Rayleighfading channel with 10 taps, uniform power delay profile, successive MMSE detection. The 3% reduce in spectral efficiency compared tosingle-carrier case (nT log2 M) is due to CP (Reproduced from [15] with permission).

concatenated SISO-OFDM-IM transmitters (Fig. 5(a)) to operate over nT ×nR MIMO frequency selectivefading channels. At the receiver of the MIMO-OFDM-IM scheme, the simultaneously transmitted OFDM-IM frames are separated and demodulated using a low-complexity MMSE detection and LLR calculationbased detector which considers the statistics of the MMSE filtered received signals. It has been shownvia extensive computer simulations that due to its improved error performance and flexible system design,MIMO-OFDM-IM can be a strong alternative to classical MIMO-OFDM, which has been included inmany current wireless standards.

In Fig. 6, the uncoded BER performance curves of the MIMO-OFDM-IM and classical V-BLASTtype MIMO-OFDM schemes are given for three MIMO configurations where the same spectral efficiencyvalues are obtained for both schemes. As observed from Fig. 6, significant signal-to-noise ratio (SNR)improvements can be obtained by the MIMO-OFDM-IM scheme compared to classical MIMO-OFDMto reach a target BER value. On the other hand, the generalization of MIMO-OFDM-IM for massiveMU-MIMO systems remains an interesting research problem towards 5G wireless networks.

Another recently proposed IM scheme, which is called generalized space-frequency index modulation(GSFIM) [17], combines OFDM-IM concept with GSM principle by exploiting both spatial and frequency(subcarrier) domains for IM. It has been shown that GSFIM scheme can also provide improvements overMIMO-OFDM in terms of achievable data rate and BER performance with ML detection for BPSK andQPSK constellations. However, the design of low complexity detector types is an open research problemfor the GSFIM scheme.

VI. CONCLUSIONS AND FUTURE WORK

IM is an up and coming concept for spectral and energy-efficient next generation wireless commu-nications systems to be employed in 5G wireless networks. IM techniques can offer low-complexity aswell as spectral and energy-efficient solutions towards the single/multi-carrier, massive MU-MIMO, andcooperative communications systems to be employed in 5G wireless networks. In this article, we havereviewed the basic principles, advantages, recent advances, and application areas of SM and OFDM-IMsystems, which are two popular applications of the IM concept. In Table II, the pros and cons of the


TABLE IIPROS AND CONS OF SEVERAL INDEX MODULATION SCHEMES

Scheme Spectralefficiency

ML detectioncomplexity

Errorperformance

Single-carriercommun.systems

SIMO Low Low LowSM Moderate Low∗ ModerateGSM Moderate Moderate∗ ModerateMA-SM High Moderate∗ ModerateESM High Low HighQSM High Low HighV-BLAST High High∗ Moderate

Multi-carriercommun.systems

OFDM Low Low LowOFDM-IM Low Moderate∗ ModerateOFDM-GIM-I Moderate High∗ ModerateOFDM-GIM-II Moderate High∗ ModerateMIMO-OFDM-IM High High∗ HighGSFIM High High ModerateV-BLAST-OFDM High Moderate∗ Moderate

∗Lower complexity near/sub-optimal detection is also possible.

reviewed IM schemes in terms of spectral efficiency, ML detection complexity, and error performanceare provided. We conclude from Table II that IM schemes can be considered as possible candidates for5G wireless networks due to the interesting trade-offs they offer among error performance, complexity,and spectral efficiency, while there are still interesting as well as challenging research problems need tobe solved in order to further improve the efficiency of IM schemes. These research challenges can besummarized as follows:

• The design of novel generalized/enhanced IM schemes with higher spectral and/or energy efficiency,lower transceiver complexity, and better error performance

• The integration of IM techniques (such as SM, GSM, ESM, QSM, and OFDM-IM) into massive MU-MIMO systems to be employed in 5G wireless networks and the design of novel uplink/downlinktransmission protocols

• The adaption of IM techniques to cooperative communications systems (such as dual/multi-hop,network-coded, multi-relay, and distributive networks)

• The investigation of the potential of IM techniques via practical implementation scenarios.


REFERENCES

[1] C.-X. Wang, F. Haider, X. Gao, X.-H. You, Y. Yang, D. Yuan, H. Aggoune, H. Haas, S. Fletcher, and E. Hepsaydir, “Cellular architectureand key technologies for 5G wireless communication networks,” IEEE Commun. Mag., vol. 52, no. 2, pp. 122–130, Feb. 2014.

[2] R. Mesleh, H. Haas, S. Sinanovic, C. W. Ahn, and S. Yun, “Spatial modulation,” IEEE Trans. Veh. Technol., vol. 57, no. 4, pp.2228–2241, Jul. 2008.

[3] J. Jeganathan, A. Ghrayeb, L. Szczecinski, and A. Ceron, “Space shift keying modulation for MIMO channels,” IEEE Trans. WirelessCommun., vol. 8, no. 7, pp. 3692–3703, Jul. 2009.

[4] E. Basar, U. Aygolu, E. Panayirci, and H. V. Poor, “Orthogonal frequency division multiplexing with index modulation,” IEEE Trans.Signal Process., vol. 61, no. 22, pp. 5536–5549, Nov. 2013.

[5] M. Di Renzo, H. Haas, A. Ghrayeb, S. Sugiura, and L. Hanzo, “Spatial modulation for generalized MIMO: Challenges, opportunities,and implementation,” Proc. of the IEEE, vol. 102, no. 1, pp. 56–103, Jan. 2014.

[6] P. Yang, M. Di Renzo, Y. Xiao, S. Li, and L. Hanzo, “Design guidelines for spatial modulation,” IEEE Commun. Surveys Tutorials,vol. 17, no. 1, pp. 6–26, First quarter 2015.

[7] A. Younis, N. Serafimovski, R. Mesleh, and H. Haas, “Generalised spatial modulation,” in (ASILOMAR), 2010 Asilomar Conf. Signals,Systems and Computers, Nov. 2010, pp. 1498–1502.

[8] J. Wang, S. Jia, and J. Song, “Generalised spatial modulation system with multiple active transmit antennas and low complexitydetection scheme,” IEEE Trans. Wireless Commun., vol. 11, no. 4, pp. 1605–1615, Apr. 2012.

[9] C.-C. Cheng, H. Sari, S. Sezginer, and Y. Su, “Enhanced spatial modulation with multiple signal constellations,” IEEE Trans. Commun.,vol. 63, no. 6, pp. 2237–2248, Jun. 2015.

[10] R. Mesleh, S. Ikki, and H. Aggoune, “Quadrature spatial modulation,” IEEE Trans. Veh. Technol., vol. 64, no. 6, pp. 2738–2742, Jun.2015.

[11] D. Basnayaka, M. Di Renzo, and H. Haas, “Massive but few active MIMO,” IEEE Trans. Veh. Technol., vol. PP, no. 99, pp. 1–17, Oct.2015.

[12] S. Narayanan, M. Chaudhry, A. Stavridis, M. Di Renzo, F. Graziosi, and H. Haas, “Multi-user spatial modulation MIMO,” in IEEEWireless Commun. Netw. Conf., Istanbul, Turkey, Apr. 2014, pp. 671–676.

[13] M. Di Renzo, S. Narayanan, F. Graziosi, and H. Haas, “Distributed spatial modulation: A cooperative diversity protocol for half–duplexrelay–aided wireless networks,” IEEE Trans. Veh. Technol., vol. PP, no. 99, pp. 1–18, Jun. 2015.

[14] M. Wen, X. Cheng, M. Ma, B. Jiao, and H. V. Poor, “On the achievable rate of OFDM with index modulation,” IEEE Trans. SignalProcess., vol. 64, no. 8, pp. 1919–1932, Apr. 2016.

[15] E. Basar, “Multiple-input multiple-output OFDM with index modulation,” IEEE Signal Process. Lett., vol. 22, no. 12, pp. 2259–2263,Dec. 2015.

[16] R. Fan, Y. Yu, and Y. Guan, “Generalization of orthogonal frequency division multiplexing with index modulation,” IEEE Trans.Wireless Commun., vol. 14, no. 10, pp. 5350 – 5359, Oct. 2015.

[17] T. Datta, H. Eshwaraiah, and A. Chockalingam, “Generalized space and frequency index modulation,” IEEE Trans. Veh. Technol.,vol. PP, no. 99, pp. 1–14, Sep. 2015.

BIOGRAPHY

Ertugrul Basar [S’09, M’13, SM’16] received his B.S. degree with high honors from Istanbul University,Istanbul, Turkey, in 2007, and his M.S. and Ph.D. degrees from Istanbul Technical University, Istanbul,Turkey, in 2009 and 2013, respectively. He spent the academic year 2011-2012 at the Department ofElectrical Engineering, Princeton University, New Jersey, USA. Currently, he is an assistant professor atIstanbul Technical University, Electronics and Communication Engineering Department and a memberof Wireless Communication Research Group. He was the recipient of Istanbul Technical University BestPh.D. Thesis Award in 2014 and won two Best Paper Awards. He is a regular Reviewer for various IEEEjournals and served as a TPC Member for several conferences. His primary research interests includeMIMO systems, index modulation, cooperative communications, OFDM, and visible light communications.

Documents

IEEE COMMUNICATIONS MAGAZINE 1 Index Modulation Techniques ... · IEEE COMMUNICATIONS MAGAZINE 1 Index Modulation Techniques for 5G Wireless Networks Ertugrul Basar, Senior Member,